o Zh@sZdZddlmZmZddlmZddlmZddlm Z m Z m Z GdddeZ dgZ d S) z Processor class for MarkupLM. )OptionalUnion) TensorType)ProcessorMixin) BatchEncodingPaddingStrategyTruncationStrategyc @seZdZdZdZdZdZ                   dded eee e fd eee e fd e e d e d e e de ede edededededede ee efdefddZddZddZeddZdS) MarkupLMProcessoraJ Constructs a MarkupLM processor which combines a MarkupLM feature extractor and a MarkupLM tokenizer into a single processor. [`MarkupLMProcessor`] offers all the functionalities you need to prepare data for the model. It first uses [`MarkupLMFeatureExtractor`] to extract nodes and corresponding xpaths from one or more HTML strings. Next, these are provided to [`MarkupLMTokenizer`] or [`MarkupLMTokenizerFast`], which turns them into token-level `input_ids`, `attention_mask`, `token_type_ids`, `xpath_tags_seq` and `xpath_subs_seq`. Args: feature_extractor (`MarkupLMFeatureExtractor`): An instance of [`MarkupLMFeatureExtractor`]. The feature extractor is a required input. tokenizer (`MarkupLMTokenizer` or `MarkupLMTokenizerFast`): An instance of [`MarkupLMTokenizer`] or [`MarkupLMTokenizerFast`]. The tokenizer is a required input. parse_html (`bool`, *optional*, defaults to `True`): Whether or not to use `MarkupLMFeatureExtractor` to parse HTML strings into nodes and corresponding xpaths. ZMarkupLMFeatureExtractor)ZMarkupLMTokenizerZMarkupLMTokenizerFastTNFradd_special_tokenspadding truncation max_lengthstridepad_to_multiple_ofreturn_token_type_idsreturn_attention_maskreturn_overflowing_tokensreturn_special_tokens_maskreturn_offsets_mapping return_lengthverbosereturn_tensorsreturncKs|jr)|dur td|dus|dus|durtd||}|d}|d}n|dur1td|dus9|dur=td|durL|jrLt|trL|g}|jdid|durW|n|d |dur`|n8dd|d |d |d |d |d| d| d| d| d| d|d|d|d|d|d||}|Sd|d |d |d |d |d| d| d| d| d| d|d|d|d|d|d||}|S)a This method first forwards the `html_strings` argument to [`~MarkupLMFeatureExtractor.__call__`]. Next, it passes the `nodes` and `xpaths` along with the additional arguments to [`~MarkupLMTokenizer.__call__`] and returns the output. Optionally, one can also provide a `text` argument which is passed along as first sequence. Please refer to the docstring of the above two methods for more information. NzDMake sure to pass HTML strings in case `parse_html` is set to `True`zUPlease don't pass nodes, xpaths nor node labels in case `parse_html` is set to `True`nodesxpathsz@You have passed HTML strings but `parse_html` is set to `False`.zIMake sure to pass nodes and xpaths in case `parse_html` is set to `False`textZ text_pair node_labelsr r r rrrrrrrrrrr) parse_html ValueErrorZfeature_extractor isinstancestr tokenizer)selfZ html_stringsrrrZ questionsr r r rrrrrrrrrrrkwargsfeaturesZencoded_inputsrr_/var/www/auris/lib/python3.10/site-packages/transformers/models/markuplm/processing_markuplm.py__call__2s!              zMarkupLMProcessor.__call__cO|jj|i|S)z This method forwards all its arguments to TrOCRTokenizer's [`~PreTrainedTokenizer.batch_decode`]. Please refer to the docstring of this method for more information. )r# batch_decoder$argsr%rrr'r*zMarkupLMProcessor.batch_decodecOr))z This method forwards all its arguments to TrOCRTokenizer's [`~PreTrainedTokenizer.decode`]. Please refer to the docstring of this method for more information. )r#decoder+rrr'r.r-zMarkupLMProcessor.decodecCs |jj}|S)N)r#model_input_names)r$Ztokenizer_input_namesrrr'r/sz#MarkupLMProcessor.model_input_names)NNNNNTFNNrNNNFFFFTN)__name__ __module__ __qualname____doc__Zfeature_extractor_classZtokenizer_classrboolrr"rr rintrrr(r*r.propertyr/rrrr'r sv       Pr N)r3typingrrZ file_utilsrZprocessing_utilsrZtokenization_utils_baserrr r __all__rrrr's   |